B1142
Title: Modeling residuals with mixtures of Gaussian with non-Gaussian components for robustness and outlier detection
Authors: Alexandra Posekany - University of Technology Vienna (Austria) [presenting]
Abstract: Outliers and systematically skewed or heavy-tailed data frequently occur in data analytical problems of many fields ranging from economics to bioinformatics. A specific notion of Bayesian robustness is robustifying the likelihood, as the backbone of the model. Constructing normally distributed likelihood models is often due to computational convenience, in the same way as classical inference with approximate normality is. Independent of sample size, data in many applied fields nowadays do not fulfil this assumption, and linear and non-linear models for regression or classification suffer from that. We aim to provide a robust estimation of parameters of the ''main part of the data'' through a normal or skewed distribution as likelihood, while simultaneously identifying the ''outlying part of the data'' represented by one or more skewed or heavy-tailed mixture components. Through the component labels and posterior weights we can identify the noisy or outlying parts of the data for filtering or inspecting the data quality.